Improved border protection: An innovative approach to detect face morphing attacks

Facial morphing attacks have shown to be a serious risk to the passport issue process, undermining border security measures. It is quite difficult to identify an attack on a morphing facial image that has been printed, scanned, and re-digitized in order to obtain a passport. This research explores a detection method that minimizes the blending-induced deterioration of images. Focused Layer-wise Relevance Propagation (FLRP) is the important part of this approach. A neural network with deep learning may be trained to identify between a real and a morphed facial image by using specific parts of the image, which this framework describes to a human inspection at the pixel level. Analysis and results from experiments indicate that this technique may significantly improve the accuracy of facial morphing attack detection. This method exhibits significant robustness and high computational efficiency when compared to the current approaches. It offers a lot of possible applications for improving the facial recognition system's security.